Nodes/ComfyUI-Prompt-Helper/Load Preset (Advanced)
ComfyUI Node

Load Preset (Advanced)

One dropdown that brings your whole style — prompts, LoRA, the lot

By Elypha·Created 2 years ago·Updated 12 months ago· 0
Load Preset (Advanced)
    • positive prompt
    • negative prompt
    • lora name
    • strength model
    • strength clip
    • lora stack
    preset

    The plain Load Preset gives you prompt blocks. This one gives you the whole style. Pick a preset from a dropdown and it hands you a positive prompt, a negative prompt, and a LoRA - all six outputs wired in, and your entire look changes with one selection. It's the node that makes "here's my style, switch to it" a one-click operation instead of a four-node edit session.

    The mechanism: presets live in JSON files in ComfyUI/custom_nodes/ComfyUI-Prompt-Helper/presets/, and each key in the file becomes a dropdown entry named filename: key. The JSON shape is:

    {
      "studio portrait": {
        "positive_prompt": "portrait of a woman, studio lighting, sharp focus",
        "negative_prompt": "blurry, lowres, deformed hands",
        "lora": { "lora_name": "portrait_style.safetensors", "weight": 0.8 }
      }
    }
    

    On execution the node parses the preset and emits six outputs: positive prompt, negative prompt, lora name, strength model, strength clip, and lora stack. The LoRA fields are the interesting part - a single lora key with a weight applies that weight to both model and CLIP strength, or you can use separate strength_model / strength_clip values for finer control. There's also a loras key (a list) if one preset should stack several LoRAs - the parsed list lands in the lora stack output, which wires straight into ComfyUI's standard LoraLoaderModelOnly node the same way a Lora Stacker would. LoRA filenames come from whatever's in your loras/ folder.

    The outputs that matter

    • positive prompt / negative prompt - wire into your two CLIP Text Encode nodes
    • lora stack - wire into LoraLoaderModelOnly to apply the preset's LoRA(s)
    • lora name, strength model, strength clip - the parsed single-LoRA values, if you'd rather load it the plain way

    Six outputs is a lot of wiring, but you do it once per workflow and from then on the dropdown is the whole control panel. Like the plain Load Preset, you create the JSON files yourself - nothing ships pre-filled - and hit ComfyUI's refresh button after editing so the dropdown reloads.

    Installing it

    Same as the rest of the pack: zero extra dependencies, no model downloads. ComfyUI Manager → ComfyUI-Prompt-Helper, or clone into custom_nodes and restart. It's under prompt_helper.

    This is arguably the most impressive node in the pack - it's a poor-man's model manager for styles. The cost is that you must define every preset by hand first, and if you're not already using LoRAs per-style, the plain Load Preset is the simpler entry point.

    Categoryprompt_helper

    Inputs (1)

    NameTypeDefaultDescription
    presetCOMBO0 options:

    Outputs (6)

    NameTypeDescription
    positive promptSTRING
    negative promptSTRING
    lora name
    strength modelFLOAT
    strength clipFLOAT
    lora stackLORA_STACK